xref: /freebsd/share/mk/meta2deps.py (revision bf7aa99a55a7692da6e55864098fd085392542b0)

#!/usr/bin/env python

from __future__ import print_function

"""
This script parses each "meta" file and extracts the
information needed to deduce build and src dependencies.

It works much the same as the original shell script, but is
*much* more efficient.

The parsing work is handled by the class MetaFile.
We only pay attention to a subset of the information in the
"meta" files.  Specifically:

'CWD'	to initialize our notion.

'C'	to track chdir(2) on a per process basis

'R'	files read are what we really care about.
	directories read, provide a clue to resolving
	subsequent relative paths.  That is if we cannot find
	them relative to 'cwd', we check relative to the last
	dir read.

'W'	files opened for write or read-write,
	for filemon V3 and earlier.

'E'	files executed.

'L'	files linked

'V'	the filemon version, this record is used as a clue
	that we have reached the interesting bit.

"""

"""
RCSid:
	$Id: meta2deps.py,v 1.45 2023/01/18 01:35:24 sjg Exp $

	Copyright (c) 2011-2020, Simon J. Gerraty
	Copyright (c) 2011-2017, Juniper Networks, Inc.
	All rights reserved.

	Redistribution and use in source and binary forms, with or without
	modification, are permitted provided that the following conditions
	are met:
	1. Redistributions of source code must retain the above copyright
	   notice, this list of conditions and the following disclaimer.
	2. Redistributions in binary form must reproduce the above copyright
	   notice, this list of conditions and the following disclaimer in the
	   documentation and/or other materials provided with the distribution.

	THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
	"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
	LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
	A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
	OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
	SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
	LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
	DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
	THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
	(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
	OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

"""

import os
import re
import sys
import stat

def resolve(path, cwd, last_dir=None, debug=0, debug_out=sys.stderr):
    """
    Return an absolute path, resolving via cwd or last_dir if needed.
    """
    if path.endswith('/.'):
        path = path[0:-2]
    if len(path) > 0 and path[0] == '/':
        if os.path.exists(path):
            return path
        if debug > 2:
            print("skipping non-existent:", path, file=debug_out)
        return None
    if path == '.':
        return cwd
    if path.startswith('./'):
        return cwd + path[1:]
    if last_dir == cwd:
        last_dir = None
    for d in [last_dir, cwd]:
        if not d:
            continue
        if path == '..':
            dw = d.split('/')
            p = '/'.join(dw[:-1])
            if not p:
                p = '/'
            return p
        p = '/'.join([d,path])
        if debug > 2:
            print("looking for:", p, end=' ', file=debug_out)
        if not os.path.exists(p):
            if debug > 2:
                print("nope", file=debug_out)
            p = None
            continue
        if debug > 2:
            print("found:", p, file=debug_out)
        return p
    return None

def cleanpath(path):
    """cleanup path without using realpath(3)"""
    if path.startswith('/'):
        r = '/'
    else:
        r = ''
    p = []
    w = path.split('/')
    for d in w:
        if not d or d == '.':
            continue
        if d == '..':
            try:
                p.pop()
                continue
            except:
                break
        p.append(d)

    return r + '/'.join(p)

def abspath(path, cwd, last_dir=None, debug=0, debug_out=sys.stderr):
    """
    Return an absolute path, resolving via cwd or last_dir if needed.
    this gets called a lot, so we try to avoid calling realpath.
    """
    rpath = resolve(path, cwd, last_dir, debug, debug_out)
    if rpath:
        path = rpath
    elif len(path) > 0 and path[0] == '/':
        return None
    if (path.find('/') < 0 or
        path.find('./') > 0 or
        path.endswith('/..')):
        path = cleanpath(path)
    return path

def sort_unique(list, cmp=None, key=None, reverse=False):
    if sys.version_info[0] == 2:
        list.sort(cmp, key, reverse)
    else:
        list.sort(reverse=reverse)
    nl = []
    le = None
    for e in list:
        if e == le:
            continue
        le = e
        nl.append(e)
    return nl

def add_trims(x):
    return ['/' + x + '/',
            '/' + x,
            x + '/',
            x]

def target_spec_exts(target_spec):
    """return a list of dirdep extensions that could match target_spec"""

    if target_spec.find(',') < 0:
        return ['.'+target_spec]
    w = target_spec.split(',')
    n = len(w)
    e = []
    while n > 0:
        e.append('.'+','.join(w[0:n]))
        n -= 1
    return e

class MetaFile:
    """class to parse meta files generated by bmake."""

    conf = None
    dirdep_re = None
    host_target = None
    srctops = []
    objroots = []
    excludes = []
    seen = {}
    obj_deps = []
    src_deps = []
    file_deps = []

    def __init__(self, name, conf={}):
        """if name is set we will parse it now.
        conf can have the follwing keys:

        SRCTOPS list of tops of the src tree(s).

        CURDIR  the src directory 'bmake' was run from.

        RELDIR  the relative path from SRCTOP to CURDIR

        MACHINE the machine we built for.
                set to 'none' if we are not cross-building.
                More specifically if machine cannot be deduced from objdirs.

        TARGET_SPEC
                Sometimes MACHINE isn't enough.

        HOST_TARGET
                when we build for the pseudo machine 'host'
                the object tree uses HOST_TARGET rather than MACHINE.

        OBJROOTS a list of the common prefix for all obj dirs it might
                end in '/' or '-'.

        DPDEPS  names an optional file to which per file dependencies
                will be appended.
                For example if 'some/path/foo.h' is read from SRCTOP
                then 'DPDEPS_some/path/foo.h +=' "RELDIR" is output.
                This can allow 'bmake' to learn all the dirs within
                the tree that depend on 'foo.h'

        EXCLUDES
                A list of paths to ignore.
                ccache(1) can otherwise be trouble.

        debug   desired debug level

        debug_out open file to send debug output to (sys.stderr)

        """

        self.name = name
        self.debug = conf.get('debug', 0)
        self.debug_out = conf.get('debug_out', sys.stderr)

        self.machine = conf.get('MACHINE', '')
        self.machine_arch = conf.get('MACHINE_ARCH', '')
        self.target_spec = conf.get('TARGET_SPEC', self.machine)
        self.exts = target_spec_exts(self.target_spec)
        self.curdir = conf.get('CURDIR')
        self.reldir = conf.get('RELDIR')
        self.dpdeps = conf.get('DPDEPS')
        self.pids = {}
        self.line = 0

        if not self.conf:
            # some of the steps below we want to do only once
            self.conf = conf
            self.host_target = conf.get('HOST_TARGET')
            for srctop in conf.get('SRCTOPS', []):
                if srctop[-1] != '/':
                    srctop += '/'
                if not srctop in self.srctops:
                    self.srctops.append(srctop)
                _srctop = os.path.realpath(srctop)
                if _srctop[-1] != '/':
                    _srctop += '/'
                if not _srctop in self.srctops:
                    self.srctops.append(_srctop)

            trim_list = add_trims(self.machine)
            if self.machine == 'host':
                trim_list += add_trims(self.host_target)
            if self.target_spec != self.machine:
                trim_list += add_trims(self.target_spec)

            for objroot in conf.get('OBJROOTS', []):
                for e in trim_list:
                    if objroot.endswith(e):
                        # this is not what we want - fix it
                        objroot = objroot[0:-len(e)]

                if objroot[-1] != '/':
                    objroot += '/'
                if not objroot in self.objroots:
                    self.objroots.append(objroot)
                    _objroot = os.path.realpath(objroot)
                    if objroot[-1] == '/':
                        _objroot += '/'
                    if not _objroot in self.objroots:
                        self.objroots.append(_objroot)

            # we want the longest match
            self.srctops.sort(reverse=True)
            self.objroots.sort(reverse=True)

            self.excludes = conf.get('EXCLUDES', [])

            if self.debug:
                print("host_target=", self.host_target, file=self.debug_out)
                print("srctops=", self.srctops, file=self.debug_out)
                print("objroots=", self.objroots, file=self.debug_out)
                print("excludes=", self.excludes, file=self.debug_out)
                print("ext_list=", self.exts, file=self.debug_out)

            self.dirdep_re = re.compile(r'([^/]+)/(.+)')

        if self.dpdeps and not self.reldir:
            if self.debug:
                print("need reldir:", end=' ', file=self.debug_out)
            if self.curdir:
                srctop = self.find_top(self.curdir, self.srctops)
                if srctop:
                    self.reldir = self.curdir.replace(srctop,'')
                    if self.debug:
                        print(self.reldir, file=self.debug_out)
            if not self.reldir:
                self.dpdeps = None      # we cannot do it?

        self.cwd = os.getcwd()          # make sure this is initialized
        self.last_dir = self.cwd

        if name:
            self.try_parse()

    def reset(self):
        """reset state if we are being passed meta files from multiple directories."""
        self.seen = {}
        self.obj_deps = []
        self.src_deps = []
        self.file_deps = []

    def dirdeps(self, sep='\n'):
        """return DIRDEPS"""
        return sep.strip() + sep.join(self.obj_deps)

    def src_dirdeps(self, sep='\n'):
        """return SRC_DIRDEPS"""
        return sep.strip() + sep.join(self.src_deps)

    def file_depends(self, out=None):
        """Append DPDEPS_${file} += ${RELDIR}
        for each file we saw, to the output file."""
        if not self.reldir:
            return None
        for f in sort_unique(self.file_deps):
            print('DPDEPS_%s += %s' % (f, self.reldir), file=out)
        # these entries provide for reverse DIRDEPS lookup
        for f in self.obj_deps:
            print('DEPDIRS_%s += %s' % (f, self.reldir), file=out)

    def seenit(self, dir):
        """rememer that we have seen dir."""
        self.seen[dir] = 1

    def add(self, list, data, clue=''):
        """add data to list if it isn't already there."""
        if data not in list:
            list.append(data)
            if self.debug:
                print("%s: %sAdd: %s" % (self.name, clue, data), file=self.debug_out)

    def find_top(self, path, list):
        """the logical tree may be split across multiple trees"""
        for top in list:
            if path.startswith(top):
                if self.debug > 2:
                    print("found in", top, file=self.debug_out)
                return top
        return None

    def find_obj(self, objroot, dir, path, input):
        """return path within objroot, taking care of .dirdep files"""
        ddep = None
        for ddepf in [path + '.dirdep', dir + '/.dirdep']:
            if not ddep and os.path.exists(ddepf):
                ddep = open(ddepf, 'r').readline().strip('# \n')
                if self.debug > 1:
                    print("found %s: %s\n" % (ddepf, ddep), file=self.debug_out)
                for e in self.exts:
                    if ddep.endswith(e):
                        ddep = ddep[0:-len(e)]
                        break

        if not ddep:
            # no .dirdeps, so remember that we've seen the raw input
            self.seenit(input)
            self.seenit(dir)
            if self.machine == 'none':
                if dir.startswith(objroot):
                    return dir.replace(objroot,'')
                return None
            m = self.dirdep_re.match(dir.replace(objroot,''))
            if m:
                ddep = m.group(2)
                dmachine = m.group(1)
                if dmachine != self.machine:
                    if not (self.machine == 'host' and
                            dmachine == self.host_target):
                        if self.debug > 2:
                            print("adding .%s to %s" % (dmachine, ddep), file=self.debug_out)
                        ddep += '.' + dmachine

        return ddep

    def try_parse(self, name=None, file=None):
        """give file and line number causing exception"""
        try:
            self.parse(name, file)
        except:
            # give a useful clue
            print('{}:{}: '.format(self.name, self.line), end=' ', file=sys.stderr)
            raise

    def parse(self, name=None, file=None):
        """A meta file looks like:

        # Meta data file "path"
        CMD "command-line"
        CWD "cwd"
        TARGET "target"
        -- command output --
        -- filemon acquired metadata --
        # buildmon version 3
        V 3
        C "pid" "cwd"
        E "pid" "path"
        F "pid" "child"
        R "pid" "path"
        W "pid" "path"
        X "pid" "status"
        D "pid" "path"
        L "pid" "src" "target"
        M "pid" "old" "new"
        S "pid" "path"
        # Bye bye

        We go to some effort to avoid processing a dependency more than once.
        Of the above record types only C,E,F,L,R,V and W are of interest.
        """

        version = 0                     # unknown
        if name:
            self.name = name;
        if file:
            f = file
            cwd = self.last_dir = self.cwd
        else:
            f = open(self.name, 'r')
        skip = True
        pid_cwd = {}
        pid_last_dir = {}
        last_pid = 0
        eof_token = False

        self.line = 0
        if self.curdir:
            self.seenit(self.curdir)    # we ignore this

        interesting = '#CEFLRVX'
        for line in f:
            self.line += 1
            # ignore anything we don't care about
            if not line[0] in interesting:
                continue
            if self.debug > 2:
                print("input:", line, end=' ', file=self.debug_out)
            w = line.split()

            if skip:
                if w[0] == 'V':
                    skip = False
                    version = int(w[1])
                    """
                    if version < 4:
                        # we cannot ignore 'W' records
                        # as they may be 'rw'
                        interesting += 'W'
                    """
                elif w[0] == 'CWD':
                    self.cwd = cwd = self.last_dir = w[1]
                    self.seenit(cwd)    # ignore this
                    if self.debug:
                        print("%s: CWD=%s" % (self.name, cwd), file=self.debug_out)
                continue

            if w[0] == '#':
                # check the file has not been truncated
                if line.find('Bye') > 0:
                    eof_token = True
                continue

            pid = int(w[1])
            if pid != last_pid:
                if last_pid:
                    pid_last_dir[last_pid] = self.last_dir
                cwd = pid_cwd.get(pid, self.cwd)
                self.last_dir = pid_last_dir.get(pid, self.cwd)
                last_pid = pid

            # process operations
            if w[0] == 'F':
                npid = int(w[2])
                pid_cwd[npid] = cwd
                pid_last_dir[npid] = cwd
                last_pid = npid
                continue
            elif w[0] == 'C':
                cwd = abspath(w[2], cwd, None, self.debug, self.debug_out)
                if not cwd:
                    cwd = w[2]
                    if self.debug > 1:
                        print("missing cwd=", cwd, file=self.debug_out)
                if cwd.endswith('/.'):
                    cwd = cwd[0:-2]
                self.last_dir = pid_last_dir[pid] = cwd
                pid_cwd[pid] = cwd
                if self.debug > 1:
                    print("cwd=", cwd, file=self.debug_out)
                continue

            if w[0] == 'X':
                try:
                    del self.pids[pid]
                except KeyError:
                    pass
                continue

            if w[2] in self.seen:
                if self.debug > 2:
                    print("seen:", w[2], file=self.debug_out)
                continue
            # file operations
            if w[0] in 'ML':
                # these are special, tread src as read and
                # target as write
                self.parse_path(w[2].strip("'"), cwd, 'R', w)
                self.parse_path(w[3].strip("'"), cwd, 'W', w)
                continue
            elif w[0] in 'ERWS':
                path = w[2]
                if w[0] == 'E':
                    self.pids[pid] = path
                elif path == '.':
                    continue
                self.parse_path(path, cwd, w[0], w)

        if version == 0:
            raise AssertionError('missing filemon data')
        if not eof_token:
            raise AssertionError('truncated filemon data')

        setid_pids = []
        # self.pids should be empty!
        for pid,path in self.pids.items():
            try:
                # no guarantee that path is still valid
                if os.stat(path).st_mode & (stat.S_ISUID|stat.S_ISGID):
                    # we do not expect anything after Exec
                    setid_pids.append(pid)
                    continue
            except:
                # we do not care why the above fails,
                # we do not want to miss the ERROR below.
                pass
            print("ERROR: missing eXit for {} pid {}".format(path, pid))
        for pid in setid_pids:
            del self.pids[pid]
        assert(len(self.pids) == 0)
        if not file:
            f.close()

    def is_src(self, base, dir, rdir):
        """is base in srctop"""
        for dir in [dir,rdir]:
            if not dir:
                continue
            path = '/'.join([dir,base])
            srctop = self.find_top(path, self.srctops)
            if srctop:
                if self.dpdeps:
                    self.add(self.file_deps, path.replace(srctop,''), 'file')
                self.add(self.src_deps, dir.replace(srctop,''), 'src')
                self.seenit(dir)
                return True
        return False

    def parse_path(self, path, cwd, op=None, w=[]):
        """look at a path for the op specified"""

        if not op:
            op = w[0]

        # we are never interested in .dirdep files as dependencies
        if path.endswith('.dirdep'):
            return
        for p in self.excludes:
            if p and path.startswith(p):
                if self.debug > 2:
                    print("exclude:", p, path, file=self.debug_out)
                return
        # we don't want to resolve the last component if it is
        # a symlink
        path = resolve(path, cwd, self.last_dir, self.debug, self.debug_out)
        if not path:
            return
        dir,base = os.path.split(path)
        if dir in self.seen:
            if self.debug > 2:
                print("seen:", dir, file=self.debug_out)
            return
        # we can have a path in an objdir which is a link
        # to the src dir, we may need to add dependencies for each
        rdir = dir
        dir = abspath(dir, cwd, self.last_dir, self.debug, self.debug_out)
        if dir:
            rdir = os.path.realpath(dir)
        else:
            dir = rdir
        if rdir == dir:
            rdir = None
        # now put path back together
        path = '/'.join([dir,base])
        if self.debug > 1:
            print("raw=%s rdir=%s dir=%s path=%s" % (w[2], rdir, dir, path), file=self.debug_out)
        if op in 'RWS':
            if path in [self.last_dir, cwd, self.cwd, self.curdir]:
                if self.debug > 1:
                    print("skipping:", path, file=self.debug_out)
                return
            if os.path.isdir(path):
                if op in 'RW':
                    self.last_dir = path;
                if self.debug > 1:
                    print("ldir=", self.last_dir, file=self.debug_out)
                return

        if op in 'ER':
            # finally, we get down to it
            if dir == self.cwd or dir == self.curdir:
                return
            if self.is_src(base, dir, rdir):
                self.seenit(w[2])
                if not rdir:
                    return

            objroot = None
            for dir in [dir,rdir]:
                if not dir:
                    continue
                objroot = self.find_top(dir, self.objroots)
                if objroot:
                    break
            if objroot:
                ddep = self.find_obj(objroot, dir, path, w[2])
                if ddep:
                    self.add(self.obj_deps, ddep, 'obj')
                    if self.dpdeps and objroot.endswith('/stage/'):
                        sp = '/'.join(path.replace(objroot,'').split('/')[1:])
                        self.add(self.file_deps, sp, 'file')
            else:
                # don't waste time looking again
                self.seenit(w[2])
                self.seenit(dir)


def main(argv, klass=MetaFile, xopts='', xoptf=None):
    """Simple driver for class MetaFile.

    Usage:
        script [options] [key=value ...] "meta" ...

    Options and key=value pairs contribute to the
    dictionary passed to MetaFile.

    -S "SRCTOP"
                add "SRCTOP" to the "SRCTOPS" list.

    -C "CURDIR"

    -O "OBJROOT"
                add "OBJROOT" to the "OBJROOTS" list.

    -m "MACHINE"

    -a "MACHINE_ARCH"

    -H "HOST_TARGET"

    -D "DPDEPS"

    -d  bumps debug level

    """
    import getopt

    # import Psyco if we can
    # it can speed things up quite a bit
    have_psyco = 0
    try:
        import psyco
        psyco.full()
        have_psyco = 1
    except:
        pass

    conf = {
        'SRCTOPS': [],
        'OBJROOTS': [],
        'EXCLUDES': [],
        }

    try:
        machine = os.environ['MACHINE']
        if machine:
            conf['MACHINE'] = machine
        machine_arch = os.environ['MACHINE_ARCH']
        if machine_arch:
            conf['MACHINE_ARCH'] = machine_arch
        srctop = os.environ['SB_SRC']
        if srctop:
            conf['SRCTOPS'].append(srctop)
        objroot = os.environ['SB_OBJROOT']
        if objroot:
            conf['OBJROOTS'].append(objroot)
    except:
        pass

    debug = 0
    output = True

    opts, args = getopt.getopt(argv[1:], 'a:dS:C:O:R:m:D:H:qT:X:' + xopts)
    for o, a in opts:
        if o == '-a':
            conf['MACHINE_ARCH'] = a
        elif o == '-d':
            debug += 1
        elif o == '-q':
            output = False
        elif o == '-H':
            conf['HOST_TARGET'] = a
        elif o == '-S':
            if a not in conf['SRCTOPS']:
                conf['SRCTOPS'].append(a)
        elif o == '-C':
            conf['CURDIR'] = a
        elif o == '-O':
            if a not in conf['OBJROOTS']:
                conf['OBJROOTS'].append(a)
        elif o == '-R':
            conf['RELDIR'] = a
        elif o == '-D':
            conf['DPDEPS'] = a
        elif o == '-m':
            conf['MACHINE'] = a
        elif o == '-T':
            conf['TARGET_SPEC'] = a
        elif o == '-X':
            if a not in conf['EXCLUDES']:
                conf['EXCLUDES'].append(a)
        elif xoptf:
            xoptf(o, a, conf)

    conf['debug'] = debug

    # get any var=val assignments
    eaten = []
    for a in args:
        if a.find('=') > 0:
            k,v = a.split('=')
            if k in ['SRCTOP','OBJROOT','SRCTOPS','OBJROOTS']:
                if k == 'SRCTOP':
                    k = 'SRCTOPS'
                elif k == 'OBJROOT':
                    k = 'OBJROOTS'
                if v not in conf[k]:
                    conf[k].append(v)
            else:
                conf[k] = v
            eaten.append(a)
            continue
        break

    for a in eaten:
        args.remove(a)

    debug_out = conf.get('debug_out', sys.stderr)

    if debug:
        print("config:", file=debug_out)
        print("psyco=", have_psyco, file=debug_out)
        for k,v in list(conf.items()):
            print("%s=%s" % (k,v), file=debug_out)

    m = None
    for a in args:
        if a.endswith('.meta'):
            if not os.path.exists(a):
                continue
            m = klass(a, conf)
        elif a.startswith('@'):
            # there can actually multiple files per line
            for line in open(a[1:]):
                for f in line.strip().split():
                    if not os.path.exists(f):
                        continue
                    m = klass(f, conf)

    if output and m:
        print(m.dirdeps())

        print(m.src_dirdeps('\nsrc:'))

        dpdeps = conf.get('DPDEPS')
        if dpdeps:
            m.file_depends(open(dpdeps, 'w'))

    return m

if __name__ == '__main__':
    try:
        main(sys.argv)
    except:
        # yes, this goes to stdout
        print("ERROR: ", sys.exc_info()[1])
        raise